LLM FenceLLM API Proxy for Teams

Secure, controlled access to multiple LLM providers

High Performance

Built in Rust on the Salvo async framework. Low memory footprint, fast startup, native high-concurrency streaming.

Auto-detect Claude Code, OpenAI Codex, and other clients. Allow or block specific clients per API key.

Three-dimensional control — Key × Client Type × Model × IP CIDR. Least-privilege access for every team member.

PreProcessor / PostProcessor / Observer architecture. Extend with a single trait implementation.

Async LLM-powered analysis of request/response content, with per-client-type preprocessing pipelines.

Static musl builds for Linux (CentOS 7+, Ubuntu 20.04+), static CRT for Windows. Single binary, simple deployment.

Documentation Notice

The project is under rapid iteration. Documentation may lag behind or be outdated. When in doubt, refer to the source code.

Under Construction

English documentation is being translated. Some pages may be incomplete. 中文文档 is the most up-to-date.